- Joined
- Nov 11, 2012
- Messages
- 1,658
Instead of my daughter always getting leftover parts for her computer, I ended up building her a dedicated gaming computer with mostly new parts. She now has a 9850X3D paired with the RX 9070 XT, instead of an i5-13600T paired with an RTX 3080m (a laptop GPU in a graphics card). Memory is set at 6000MHz CL30 @ 1.4V's, and SOC voltage undervolted at 1.1V's instead of at Auto.
While losing some light single-threaded performance boost, the gains of a manual all-core overclock seemed better. Basically, all cores at 5.450MHz at 1.21V's with LLC Level 6 and FCLK at 2000MHz. Temps are better, not going past 74C, and the highest voltage I saw was 1.3202V's. Really simple overclock, with good temps.
Old Comparison scores vs stock: I knew going in that the benchmark scores were going to be lower, but what I'm counting on is better FPS in games.
Y-Cruncher stock = 126,073 seconds
Y-Cruncher OC = 133,423 seconds
CPU-Z stock Single Thread = 889.4
CPU-Z OC Single Thread = 864.2
CPU-Z stock Multi Thread = 8968.9
CPU-Z OC Multi Thread = 9148.5
Geekbench6 stock Single-Core = 3535
Geekbench6 OC Single-Core = 3316
Geekbench6 stock Multi-Core = 19715
Geekbench6 OC Multi-Core = 19079
2K high native:
Cyberpunk2077 Average FPS stock = 154.78
Cyberpunk2077 Average FPS OC = 156.05
Cyberpunk2077 Minimum FPS stock = 129.73
Cyberpunk2077 Minimum FPS OC = 132.76
On hindsight, I should have also tested at 1080P, but since she'll be gaming at 2K, it seemed logical just to test how she'd be gaming.
It performs better at 2K against an i7-14700T with DDR4 memory, but not enough to warrant the price differential:
Cyberpunk2077 Average FPS = 149.00 (-4.62% slower vs OC 9850X3D)
Cyberpunk2077 Minimum FPS = 123.77 (-7.01% slower vs OC 9850X3D)

Pic of build:

Update:
I've been reading this as well (a Kingston article). Theoretically, 8000MHz CL38 at Gear 2 (which I was able to do), would have faster latency than 6000MHz CL30 at Gear 1.
View attachment 2769278
View attachment 2769276
View attachment 2769277
As far as gaming, I got a slight edge in Min FPS at 6000 CL30 at Gear 1, but a slight edge in Max FPS at 8000MHz CL38 at Gear 2. I can't tell any perceptible difference, everything is snappy as hell. I'll post updated benchmarks soon.
While losing some light single-threaded performance boost, the gains of a manual all-core overclock seemed better. Basically, all cores at 5.450MHz at 1.21V's with LLC Level 6 and FCLK at 2000MHz. Temps are better, not going past 74C, and the highest voltage I saw was 1.3202V's. Really simple overclock, with good temps.
Old Comparison scores vs stock: I knew going in that the benchmark scores were going to be lower, but what I'm counting on is better FPS in games.
Y-Cruncher stock = 126,073 seconds
Y-Cruncher OC = 133,423 seconds
CPU-Z stock Single Thread = 889.4
CPU-Z OC Single Thread = 864.2
CPU-Z stock Multi Thread = 8968.9
CPU-Z OC Multi Thread = 9148.5
Geekbench6 stock Single-Core = 3535
Geekbench6 OC Single-Core = 3316
Geekbench6 stock Multi-Core = 19715
Geekbench6 OC Multi-Core = 19079
2K high native:
Cyberpunk2077 Average FPS stock = 154.78
Cyberpunk2077 Average FPS OC = 156.05
Cyberpunk2077 Minimum FPS stock = 129.73
Cyberpunk2077 Minimum FPS OC = 132.76
On hindsight, I should have also tested at 1080P, but since she'll be gaming at 2K, it seemed logical just to test how she'd be gaming.
It performs better at 2K against an i7-14700T with DDR4 memory, but not enough to warrant the price differential:
Cyberpunk2077 Average FPS = 149.00 (-4.62% slower vs OC 9850X3D)
Cyberpunk2077 Minimum FPS = 123.77 (-7.01% slower vs OC 9850X3D)
Pic of build:
Update:
I've been reading this as well (a Kingston article). Theoretically, 8000MHz CL38 at Gear 2 (which I was able to do), would have faster latency than 6000MHz CL30 at Gear 1.
View attachment 2769278
View attachment 2769276
View attachment 2769277
As far as gaming, I got a slight edge in Min FPS at 6000 CL30 at Gear 1, but a slight edge in Max FPS at 8000MHz CL38 at Gear 2. I can't tell any perceptible difference, everything is snappy as hell. I'll post updated benchmarks soon.
Last edited: